gemm: General matrix multiply: C = AB + C

\( C = \alpha \;op(A) \;op(B) + \beta C \) More...

Functions
int	plasma_cgemm (plasma_enum_t transa, plasma_enum_t transb, int m, int n, int k, plasma_complex32_t alpha, plasma_complex32_t *pA, int lda, plasma_complex32_t *pB, int ldb, plasma_complex32_t beta, plasma_complex32_t *pC, int ldc)
void	plasma_omp_cgemm (plasma_enum_t transa, plasma_enum_t transb, plasma_complex32_t alpha, plasma_desc_t A, plasma_desc_t B, plasma_complex32_t beta, plasma_desc_t C, plasma_sequence_t *sequence, plasma_request_t *request)
int	plasma_dgemm (plasma_enum_t transa, plasma_enum_t transb, int m, int n, int k, double alpha, double *pA, int lda, double *pB, int ldb, double beta, double *pC, int ldc)
void	plasma_omp_dgemm (plasma_enum_t transa, plasma_enum_t transb, double alpha, plasma_desc_t A, plasma_desc_t B, double beta, plasma_desc_t C, plasma_sequence_t *sequence, plasma_request_t *request)
int	plasma_sgemm (plasma_enum_t transa, plasma_enum_t transb, int m, int n, int k, float alpha, float *pA, int lda, float *pB, int ldb, float beta, float *pC, int ldc)
void	plasma_omp_sgemm (plasma_enum_t transa, plasma_enum_t transb, float alpha, plasma_desc_t A, plasma_desc_t B, float beta, plasma_desc_t C, plasma_sequence_t *sequence, plasma_request_t *request)
int	plasma_zgemm (plasma_enum_t transa, plasma_enum_t transb, int m, int n, int k, plasma_complex64_t alpha, plasma_complex64_t *pA, int lda, plasma_complex64_t *pB, int ldb, plasma_complex64_t beta, plasma_complex64_t *pC, int ldc)
void	plasma_omp_zgemm (plasma_enum_t transa, plasma_enum_t transb, plasma_complex64_t alpha, plasma_desc_t A, plasma_desc_t B, plasma_complex64_t beta, plasma_desc_t C, plasma_sequence_t *sequence, plasma_request_t *request)

Detailed Description

\( C = \alpha \;op(A) \;op(B) + \beta C \)

Function Documentation

int plasma_cgemm	(	plasma_enum_t	transa,
		plasma_enum_t	transb,
		int	m,
		int	n,
		int	k,
		plasma_complex32_t	alpha,
		plasma_complex32_t *	pA,
		int	lda,
		plasma_complex32_t *	pB,
		int	ldb,
		plasma_complex32_t	beta,
		plasma_complex32_t *	pC,
		int	ldc
	)

Performs one of the matrix-matrix operations

\[ C = \alpha [op( A )\times op( B )] + \beta C, \]

where op( X ) is one of:

\[ op( X ) = X, \]

\[ op( X ) = X^T, \]

\[ op( X ) = X^H, \]

alpha and beta are scalars, and A, B and C are matrices, with op( A ) an m-by-k matrix, op( B ) a k-by-n matrix and C an m-by-n matrix.

Parameters

[in]	transa	PlasmaNoTrans: A is not transposed, PlasmaTrans: A is transposed, PlasmaConjTrans: A is conjugate transposed.
[in]	transb	PlasmaNoTrans: B is not transposed, PlasmaTrans: B is transposed, PlasmaConjTrans: B is conjugate transposed.
[in]	m	The number of rows of the matrix op( A ) and of the matrix C. m >= 0.
[in]	n	The number of columns of the matrix op( B ) and of the matrix C. n >= 0.
[in]	k	The number of columns of the matrix op( A ) and the number of rows of the matrix op( B ). k >= 0.
[in]	alpha	The scalar alpha.
[in]	pA	An lda-by-ka matrix, where ka is k when transa = PlasmaNoTrans, and is m otherwise.
[in]	lda	The leading dimension of the array A. When transa = PlasmaNoTrans, lda >= max(1,m), otherwise, lda >= max(1,k).
[in]	pB	An ldb-by-kb matrix, where kb is n when transb = PlasmaNoTrans, and is k otherwise.
[in]	ldb	The leading dimension of the array B. When transb = PlasmaNoTrans, ldb >= max(1,k), otherwise, ldb >= max(1,n).
[in]	beta	The scalar beta.
[in,out]	pC	An ldc-by-n matrix. On exit, the array is overwritten by the m-by-n matrix ( alpha*op( A )*op( B ) + beta*C ).
[in]	ldc	The leading dimension of the array C. ldc >= max(1,m).

Return values

PlasmaSuccess

successful exit

See also

plasma_omp_cgemm

plasma_cgemm

plasma_dgemm

plasma_sgemm

void plasma_omp_cgemm	(	plasma_enum_t	transa,
		plasma_enum_t	transb,
		plasma_complex32_t	alpha,
		plasma_desc_t	A,
		plasma_desc_t	B,
		plasma_complex32_t	beta,
		plasma_desc_t	C,
		plasma_sequence_t *	sequence,
		plasma_request_t *	request
	)

Performs matrix multiplication. Non-blocking tile version of plasma_cgemm(). May return before the computation is finished. Operates on matrices stored by tiles. All matrices are passed through descriptors. All dimensions are taken from the descriptors. Allows for pipelining of operations at runtime.

Parameters

[in]	transa	PlasmaNoTrans: A is not transposed, PlasmaTrans: A is transposed, PlasmaConjTrans: A is conjugate transposed.
[in]	transb	PlasmaNoTrans: B is not transposed, PlasmaTrans: B is transposed, PlasmaConjTrans: B is conjugate transposed.
[in]	alpha	The scalar alpha.
[in]	A	Descriptor of matrix A.
[in]	B	Descriptor of matrix B.
[in]	beta	The scalar beta.
[in,out]	C	Descriptor of matrix C.
[in]	sequence	Identifies the sequence of function calls that this call belongs to (for completion checks and exception handling purposes). Check the sequence->status for errors.
[out]	request	Identifies this function call (for exception handling purposes).

Return values

void	Errors are returned by setting sequence->status and request->status to error values. The sequence->status and request->status should never be set to PlasmaSuccess (the initial values) since another async call may be setting a failure value at the same time.

See also

plasma_cgemm

plasma_omp_cgemm

plasma_omp_dgemm

plasma_omp_sgemm

int plasma_dgemm	(	plasma_enum_t	transa,
		plasma_enum_t	transb,
		int	m,
		int	n,
		int	k,
		double	alpha,
		double *	pA,
		int	lda,
		double *	pB,
		int	ldb,
		double	beta,
		double *	pC,
		int	ldc
	)

Performs one of the matrix-matrix operations

\[ C = \alpha [op( A )\times op( B )] + \beta C, \]

where op( X ) is one of:

\[ op( X ) = X, \]

\[ op( X ) = X^T, \]

\[ op( X ) = X^T, \]

alpha and beta are scalars, and A, B and C are matrices, with op( A ) an m-by-k matrix, op( B ) a k-by-n matrix and C an m-by-n matrix.

Parameters

[in]	transa	PlasmaNoTrans: A is not transposed, PlasmaTrans: A is transposed, PlasmaConjTrans: A is conjugate transposed.
[in]	transb	PlasmaNoTrans: B is not transposed, PlasmaTrans: B is transposed, PlasmaConjTrans: B is conjugate transposed.
[in]	m	The number of rows of the matrix op( A ) and of the matrix C. m >= 0.
[in]	n	The number of columns of the matrix op( B ) and of the matrix C. n >= 0.
[in]	k	The number of columns of the matrix op( A ) and the number of rows of the matrix op( B ). k >= 0.
[in]	alpha	The scalar alpha.
[in]	pA	An lda-by-ka matrix, where ka is k when transa = PlasmaNoTrans, and is m otherwise.
[in]	lda	The leading dimension of the array A. When transa = PlasmaNoTrans, lda >= max(1,m), otherwise, lda >= max(1,k).
[in]	pB	An ldb-by-kb matrix, where kb is n when transb = PlasmaNoTrans, and is k otherwise.
[in]	ldb	The leading dimension of the array B. When transb = PlasmaNoTrans, ldb >= max(1,k), otherwise, ldb >= max(1,n).
[in]	beta	The scalar beta.
[in,out]	pC	An ldc-by-n matrix. On exit, the array is overwritten by the m-by-n matrix ( alpha*op( A )*op( B ) + beta*C ).
[in]	ldc	The leading dimension of the array C. ldc >= max(1,m).

Return values

PlasmaSuccess

successful exit

See also

plasma_omp_dgemm

plasma_cgemm

plasma_dgemm

plasma_sgemm

void plasma_omp_dgemm	(	plasma_enum_t	transa,
		plasma_enum_t	transb,
		double	alpha,
		plasma_desc_t	A,
		plasma_desc_t	B,
		double	beta,
		plasma_desc_t	C,
		plasma_sequence_t *	sequence,
		plasma_request_t *	request
	)

Performs matrix multiplication. Non-blocking tile version of plasma_dgemm(). May return before the computation is finished. Operates on matrices stored by tiles. All matrices are passed through descriptors. All dimensions are taken from the descriptors. Allows for pipelining of operations at runtime.

Parameters

[in]	transa	PlasmaNoTrans: A is not transposed, PlasmaTrans: A is transposed, PlasmaConjTrans: A is conjugate transposed.
[in]	transb	PlasmaNoTrans: B is not transposed, PlasmaTrans: B is transposed, PlasmaConjTrans: B is conjugate transposed.
[in]	alpha	The scalar alpha.
[in]	A	Descriptor of matrix A.
[in]	B	Descriptor of matrix B.
[in]	beta	The scalar beta.
[in,out]	C	Descriptor of matrix C.
[in]	sequence	Identifies the sequence of function calls that this call belongs to (for completion checks and exception handling purposes). Check the sequence->status for errors.
[out]	request	Identifies this function call (for exception handling purposes).

Return values

void	Errors are returned by setting sequence->status and request->status to error values. The sequence->status and request->status should never be set to PlasmaSuccess (the initial values) since another async call may be setting a failure value at the same time.

See also

plasma_dgemm

plasma_omp_cgemm

plasma_omp_dgemm

plasma_omp_sgemm

int plasma_sgemm	(	plasma_enum_t	transa,
		plasma_enum_t	transb,
		int	m,
		int	n,
		int	k,
		float	alpha,
		float *	pA,
		int	lda,
		float *	pB,
		int	ldb,
		float	beta,
		float *	pC,
		int	ldc
	)

Performs one of the matrix-matrix operations

\[ C = \alpha [op( A )\times op( B )] + \beta C, \]

where op( X ) is one of:

\[ op( X ) = X, \]

\[ op( X ) = X^T, \]

\[ op( X ) = X^T, \]

alpha and beta are scalars, and A, B and C are matrices, with op( A ) an m-by-k matrix, op( B ) a k-by-n matrix and C an m-by-n matrix.

Parameters

[in]	transa	PlasmaNoTrans: A is not transposed, PlasmaTrans: A is transposed, PlasmaConjTrans: A is conjugate transposed.
[in]	transb	PlasmaNoTrans: B is not transposed, PlasmaTrans: B is transposed, PlasmaConjTrans: B is conjugate transposed.
[in]	m	The number of rows of the matrix op( A ) and of the matrix C. m >= 0.
[in]	n	The number of columns of the matrix op( B ) and of the matrix C. n >= 0.
[in]	k	The number of columns of the matrix op( A ) and the number of rows of the matrix op( B ). k >= 0.
[in]	alpha	The scalar alpha.
[in]	pA	An lda-by-ka matrix, where ka is k when transa = PlasmaNoTrans, and is m otherwise.
[in]	lda	The leading dimension of the array A. When transa = PlasmaNoTrans, lda >= max(1,m), otherwise, lda >= max(1,k).
[in]	pB	An ldb-by-kb matrix, where kb is n when transb = PlasmaNoTrans, and is k otherwise.
[in]	ldb	The leading dimension of the array B. When transb = PlasmaNoTrans, ldb >= max(1,k), otherwise, ldb >= max(1,n).
[in]	beta	The scalar beta.
[in,out]	pC	An ldc-by-n matrix. On exit, the array is overwritten by the m-by-n matrix ( alpha*op( A )*op( B ) + beta*C ).
[in]	ldc	The leading dimension of the array C. ldc >= max(1,m).

Return values

PlasmaSuccess

successful exit

See also

plasma_omp_sgemm

plasma_cgemm

plasma_dgemm

plasma_sgemm

void plasma_omp_sgemm	(	plasma_enum_t	transa,
		plasma_enum_t	transb,
		float	alpha,
		plasma_desc_t	A,
		plasma_desc_t	B,
		float	beta,
		plasma_desc_t	C,
		plasma_sequence_t *	sequence,
		plasma_request_t *	request
	)

Performs matrix multiplication. Non-blocking tile version of plasma_sgemm(). May return before the computation is finished. Operates on matrices stored by tiles. All matrices are passed through descriptors. All dimensions are taken from the descriptors. Allows for pipelining of operations at runtime.

Parameters

[in]	transa	PlasmaNoTrans: A is not transposed, PlasmaTrans: A is transposed, PlasmaConjTrans: A is conjugate transposed.
[in]	transb	PlasmaNoTrans: B is not transposed, PlasmaTrans: B is transposed, PlasmaConjTrans: B is conjugate transposed.
[in]	alpha	The scalar alpha.
[in]	A	Descriptor of matrix A.
[in]	B	Descriptor of matrix B.
[in]	beta	The scalar beta.
[in,out]	C	Descriptor of matrix C.
[in]	sequence	Identifies the sequence of function calls that this call belongs to (for completion checks and exception handling purposes). Check the sequence->status for errors.
[out]	request	Identifies this function call (for exception handling purposes).

Return values

void	Errors are returned by setting sequence->status and request->status to error values. The sequence->status and request->status should never be set to PlasmaSuccess (the initial values) since another async call may be setting a failure value at the same time.

See also

plasma_sgemm

plasma_omp_cgemm

plasma_omp_dgemm

plasma_omp_sgemm

int plasma_zgemm	(	plasma_enum_t	transa,
		plasma_enum_t	transb,
		int	m,
		int	n,
		int	k,
		plasma_complex64_t	alpha,
		plasma_complex64_t *	pA,
		int	lda,
		plasma_complex64_t *	pB,
		int	ldb,
		plasma_complex64_t	beta,
		plasma_complex64_t *	pC,
		int	ldc
	)

Performs one of the matrix-matrix operations

\[ C = \alpha [op( A )\times op( B )] + \beta C, \]

where op( X ) is one of:

\[ op( X ) = X, \]

\[ op( X ) = X^T, \]

\[ op( X ) = X^H, \]

alpha and beta are scalars, and A, B and C are matrices, with op( A ) an m-by-k matrix, op( B ) a k-by-n matrix and C an m-by-n matrix.

Parameters

[in]	transa	PlasmaNoTrans: A is not transposed, PlasmaTrans: A is transposed, PlasmaConjTrans: A is conjugate transposed.
[in]	transb	PlasmaNoTrans: B is not transposed, PlasmaTrans: B is transposed, PlasmaConjTrans: B is conjugate transposed.
[in]	m	The number of rows of the matrix op( A ) and of the matrix C. m >= 0.
[in]	n	The number of columns of the matrix op( B ) and of the matrix C. n >= 0.
[in]	k	The number of columns of the matrix op( A ) and the number of rows of the matrix op( B ). k >= 0.
[in]	alpha	The scalar alpha.
[in]	pA	An lda-by-ka matrix, where ka is k when transa = PlasmaNoTrans, and is m otherwise.
[in]	lda	The leading dimension of the array A. When transa = PlasmaNoTrans, lda >= max(1,m), otherwise, lda >= max(1,k).
[in]	pB	An ldb-by-kb matrix, where kb is n when transb = PlasmaNoTrans, and is k otherwise.
[in]	ldb	The leading dimension of the array B. When transb = PlasmaNoTrans, ldb >= max(1,k), otherwise, ldb >= max(1,n).
[in]	beta	The scalar beta.
[in,out]	pC	An ldc-by-n matrix. On exit, the array is overwritten by the m-by-n matrix ( alpha*op( A )*op( B ) + beta*C ).
[in]	ldc	The leading dimension of the array C. ldc >= max(1,m).

Return values

PlasmaSuccess

successful exit

See also

plasma_omp_zgemm

plasma_cgemm

plasma_dgemm

plasma_sgemm

void plasma_omp_zgemm	(	plasma_enum_t	transa,
		plasma_enum_t	transb,
		plasma_complex64_t	alpha,
		plasma_desc_t	A,
		plasma_desc_t	B,
		plasma_complex64_t	beta,
		plasma_desc_t	C,
		plasma_sequence_t *	sequence,
		plasma_request_t *	request
	)

Performs matrix multiplication. Non-blocking tile version of plasma_zgemm(). May return before the computation is finished. Operates on matrices stored by tiles. All matrices are passed through descriptors. All dimensions are taken from the descriptors. Allows for pipelining of operations at runtime.

Parameters

[in]	transa	PlasmaNoTrans: A is not transposed, PlasmaTrans: A is transposed, PlasmaConjTrans: A is conjugate transposed.
[in]	transb	PlasmaNoTrans: B is not transposed, PlasmaTrans: B is transposed, PlasmaConjTrans: B is conjugate transposed.
[in]	alpha	The scalar alpha.
[in]	A	Descriptor of matrix A.
[in]	B	Descriptor of matrix B.
[in]	beta	The scalar beta.
[in,out]	C	Descriptor of matrix C.
[in]	sequence	Identifies the sequence of function calls that this call belongs to (for completion checks and exception handling purposes). Check the sequence->status for errors.
[out]	request	Identifies this function call (for exception handling purposes).

Return values

void	Errors are returned by setting sequence->status and request->status to error values. The sequence->status and request->status should never be set to PlasmaSuccess (the initial values) since another async call may be setting a failure value at the same time.

See also

plasma_zgemm

plasma_omp_cgemm

plasma_omp_dgemm

plasma_omp_sgemm